On 10/08/2018 23:13, Junaid Shahid wrote: >> Perhaps you can write a standalone test (not using kvm-unit-tests) >> you can place the PDPTEs in a place where the read will cause a >> vmentry failure? > > It wouldn't cause a vmentry failure because the paths that actually > load the PDPTEs into the VMCS already have the is_paging() check. > These two paths are just about kvm itself reading them unnecessarily > (and then doing further work when it thinks that the PDPTEs have > changed). Then the patch is okay, thanks for the clarification! Paolo